The Dirac Equation

January 19, 2011

0.1

Introduction

In non-relativistic quantum mechanics, wave functions are descibed by the time-dependent Schrodinger equation : 1 2 +V =i 2m t
2

(1)

p This is really just energy conservation ( kinetic energy ( 2m ) plus potential energy (V) equals total energy (E)) with the momentum and energy terms replaced by their operator equivalents

p i; E i

(2)

In relativistic quantum theory, the energy-momentum conservation equation is E 2 p2 = m21 . Proceeding with the same replacements, we can derive the Klein-Gordon equation : E 2 p2 m2 In covariant notation this is m2 = 0 (4) 2 + 2 m2 = 0 t2 (3)

Suppose is solution to the Klein-Gordon equation. Multiplying it by i we get 2 i 2 i 2 + i m2 = 0 (5) t Taking the complex conjugate of the Klein-Gordon equation and multiplying by i we get 2 (6) i 2 i2 + im2 = 0 t If we subtract the second from the rst we obtain i t t t
1

+ [i ( )] = 0

(7)

We are working in the standard particle physics units where = c = 1 h

This has the form of an equation of continuity +j=0 t with a probability density dened by = i t t (9) (8)

and a probability density current dened by j = i ( ) (10)

Now, suppose a solution to the Klein-Gordon equation is a free particle with energy E and momentum p = Neip x

(11)

Substitution of this solution into the equation for the probability density yields = 2E|N|2 (12) The probability density is proportional to the energy of the particle. Now, why is this a problem? If you substitute the free particle solution into the Klein-Gordon equation you get, unsurpisingly, the relation E 2 p2 = m2 so the energy of the particle could be E = p2 + m2 (14) (13)

The fact that you have negative energy solutions is not that much of a problem. What is a problem is that the probability density is proportional to E. This implies a possibility for negative probability densities...which dont exist. This (and some others) problem drove Dirac to think about another equation of motion. His starting point was to try to factorise the energy momentum relation. In covariant formalism E 2 p2 m2 p p m2 2 (15)

where p is the 4-momentum : (E, px , py , pz ). Dirac tried to write p p m2 = ( p + m)( p m) Expanding the right-hand side ( p + m)( p m) = p p m2 + m p m p (17) (16)

This should be equal to p p m2 , so we need to get rid of the terms that are linear in p. We can do this just be choosing = . Then p p m2 = p p m2 (18)

Now, since , runs from 0 to 3, the right hand side can be fully expressed by p p m2 = + + + This must be equal to p p m2 = p2 p2 p2 p2 m2 0 1 2 3 (23) ( 0 )2 p2 + ( 1 )2 p2 + ( 2 )2 p2 + ( 3 )2 p2 0 1 2 3 ( 0 1 + 1 0 )p0 p1 + ( 0 2 + 2 0 )p0 p2 ( 0 3 + 3 0 )p0 p3 + ( 1 2 + 2 1 )p1 p2 ( 1 3 + 3 1 )p1 p3 + ( 2 3 + 3 2 )p2 p3 m2 (19) (20) (21) (22)

One could make the squared terms equal by choosing ( 0 )2 = 1 and ( 1 )2 = ( 2 )2 = ( 3 )2 =-1, but this would not eliminate the cross-terms. Dirac realised that what you needed was something which anticommuted : i.e. + = 0 if = . He realised that the factors must be 4x4 matrices, not just numbers, which satised the anticommutation relation { , } = 2g where g 1 0 0 0 0 1 0 0 = 0 0 1 0 0 0 0 1 (24)

(25)

We also dene, for later use, another matrix called 5 = i 0 1 2 3 . This has the properties (you should check for yourself) that 3

( 5 )2 = 1, { 5 , } = 0

(26)

We will want to take the Hermitian conjugate of these matrices at various times. The Hermitian conjugate of each matrix is 0 = 0 5 = 5 = 0 0 = for = 0 (27)

If the matrices satisfy this algebra, then you can factor the energy momentum conservation equation p p m2 = ( p + m)( p m) = 0 (28)

The Dirac equation is one of the two factors, and is conventionally taken to be p m = 0 (29)

Making the standard substitution, p i we then have the usual covariant form of the Dirac equation (i m) = 0 (30)

where = ( t , x , y , z ), m is the particle mass and the matrices are a set of 4-dimensional matrices. Since these are matrices, is a 4-element column matrix called a bi-spinor. This bi-spinor is not a 4-vector and doesnt transform like one.

0.2

The Bjorken-Drell convention

Any set of four 4x4 matrices that obey the algebra above will work. The one we normally use includes the Pauli spin matrices. Recall that each component of the spin operator S for spin 1/2 particles is dened by its own Pauli spin matrix : 1 1 0 1 Sx = 1 = 2 2 1 0 1 1 1 0 Sz = 3 = 2 2 0 1 (31) In terms of the Pauli spin matrices, the matrices have the form 0 i i 0 4 1 1 Sy = 2 = 2 2

0 =

1 0 0 1

0 0

5 =

0 1 1 0

(32)

Each element of the matrices in Equations 32 are 2x2 matrices. 1 denotes the 2 x 2 unit matrix, and 0 denotes the 2 x 2 null matrix.

0.3

The Probability Density and Current

In order to understand the probability density and probability ow we will want to derive an equation of continuity for the probability. The rst step is to write the Dirac equation out longhand : i 0 + i 1 + i 2 + i 3 m = 0 t x y z (33)

We want to take the Hermitian conjugate of this : [i 0 + i 1 + i 2 + i 3 m] t x y z (34)

Now, we must remember that the are matrices and that is a 4component column vector. Thus 1 0 0 0 1 0 = [ 0 0 1 0 0 0 1 1 t 2 0 t = 3 0 t 4 0 t =
1 t 2 t

[ 0

] t

0 t

1 0 t 0 2 t ] 0 3 t 4 1 t 0 0 0 1 0 0 0 1 0 0 0 1 1 0 0 0 0 0 3 4 0 1 0 0 1 0 t t 0 0 0 1

(35)

(36)

(37)

(38) 5

Using the Hermitian conjugation properties of the matrices dened in the previous section we can write Equation 34 as i 0 1 2 3 i i i m t x y z (39)

which, as = for = 0 means we can write this as i 0 i ( 1 ) i ( 2 ) i ( 3 ) m t x y z (40)

We have a problem now - this is no longer covariant. That is, I would like to write this in the form (i m) where the derivative now operates to the left. I cant because the minus signs on the spatial components coming from the spoils the scalar product. To deal with this we have to multiply the equation on the right by 0 , since we can ip the sign of the matrices using the relationship 0 = 0 . Thus 0 0 i ( 1 0 ) i ( 2 0 ) i ( 3 0 ) m 0 (41) t x y z

or 0 0 0 1 0 2 0 3 i ( ) i ( ) i ( ) m 0 t x y z

(42)

Dening the adjoint spinor = 0 , we can rewrite this equation as i 0 i ( 1 ) i ( 2 ) i ( 3 ) m t x y z (43)

and nally we get the adjoint Dirac equation (i + m) = 0 Now, we multiply the Dirac equation on the left by : (i m) = 0 (45) (44)

and the adjoint Dirac equation on the right by : (i + m) = 0 and we add them together, which eliminates the two mass terms ( ) + ( ) = 0 or, more succinctly, ( ) = 0 . If we now identify the bit in the brackets as the 4-current : j = (, j) (49) (48) (47) (46)

where is the probability density and j is the probability current, we can write the conservation equation as j = 0 with j = (50)

which is the covariant form for an equation of continuity. The probability density is then just = 0 = 0 0 = and the probability 3current is j = 0 . This current is the same one which appears in the Feynman diagrams. It is called a Vector current, and is the current responsible for the electromagnetic interaction.

i e q e i
7

f e e f

For the interaction in Figure 0.3, with two electromagnetic interactions, the matrix element is then M= e2 1 [f i ] 2 [f i ] 4 q (51)

0.4

Examples

Now lets look at some solutions to the Dirac Equation. The rst one we will look at is for a particle at rest.

0.4.1

Particle at rest
(p) = eix p u(p)

Suppose the fermion wavefunction is a plane wave. We can write this as (52)

where p = (E, px , py , pz ) and x = (t, x, y, z) and so ix p = i(Et x p). This is just the phase for an oscillating plane wave. For a particle at rest, the momentum term disappears : ix p = i(Et). Furthur, since the momentum is zero, the spatial derivatives must be zero : = 0. The Dirac equation therefore reads x,y,z i 0 or i 0 (iE)u(p) mu(p) = 0 which gives us E 0 u(p) = mu(p) Now, u(p) is a 4-component matrix in Equation 55, we get 1 0 0 1 E 0 0 0 0 (55) bispinor, so properly expanding the gamma u1 0 0 u2 0 0 = m u3 1 0 u4 0 1 (54) m = 0 t (53)

(56)

This is an eigenvalue equation. There are four independent solutions. Two with energy E = m and two with E = m. The solutions are just 0 0 0 1 0 0 1 0 (57) u1 = u2 = u3 = u4 = 0 1 0 0 1 0 0 0 with eigenvalues m, m, -m and -m respectively. The rst two solutions can be interpreted as positive energy particle solutions with spin up and spin down. One can see this be checking if the function is an eigenfunction of the spin matrix : Sz . 1 1 0 0 0 0 1 0 0 0 1 1 = u1 Sz u 1 = (58) 2 0 0 1 0 0 2 0 0 0 0 1 and similarly for u2 . Note that we have yet to normalise these solutions. We wont either for the purposes of this discussion. What about these negative energy solutions? We are required to keep them (we cant just say that they are unphysical and throw them away) since quantum mechanics requires a complete set of states. What happens if we allow the particle to have momentum?

0.4.2

Particle in Motion
(p) = eix p u(p)

Again, lets look for a plane wave solution (59)

Plugging this into the Dirac equation means that we can replace the by p and remove the oscillatory term to obtain the momentum-space version of the Dirac Equation ( p m)u(p) = 0 (60)

Notice that this is now purely algebraic and can be easily(!) solved p m = E 0 px 1 py 2 pz 3 m 1 0 0 1 0 = E pm 0 1 0 0 1 = (E m) p p (E + m) (61) (62) (63)

where each element in this 2x2 matrix is actually a 2x2 matrix itself. In this spirit, lets write the 4-component bispinor solution as 2-component vector u= then the Dirac Equation implies that ( p m)u(p) = 0 = (E m) p p (E + m) uA uB = 0 (65) 0 uA uB (64)

leading two coupled simultaneous equations ( p)uB = (E m)uA ( p)uA = (E + m)uB Now, lets expand that ( p) : ( p) = = 1 0 0 1 0 1 py + px + 0 1 i 0 1 0 pz px ipy px + ipy pz (68) (69) (66) (67)

and right back to the Dirac Equation ( p)uB = (E m)uA ( p)uA = (E + m)uB gives uB = ( p) uA E+m 1 pz px ipy uA = pz E + m px + ipy 10 (72) (73) (70) (71)

Now, we just need to make choices for the form of uA . Lets make the obvious choice and remember that uA is a 2-component spinor so we need to specify two solutions: uA = These give 1 0 u1 = N1 pz E+m
px +ipy E+m

1 0

or uA =

0 1

(74)

0 1 and u2 = N2 px ipy
E+m pz E+m

(75)

where N1 and N2 are normalisation factors we wont go into. Note that, if p = 0 these correspond to the E > 0 solutions we found for a particle at rest, which is good. 1 0 Repeating this for uB = and uB = which gives solutions u3 0 1 and u4 pz px ipy px +ipy u3 = N3 Em 1 0
Em

pz u4 = N4 Em u3 = N3 0 1 E+m px +ipy pz 0 1 E+m E+m (77) All of which, if put back into the Dirac Equation, yields : E 2 = p2 + m2 as you might expect. Comparing with the p = 0 case we can identify u1 and u2 as positive energy solutions with energy E = p2 + m2 , and u3 , u4 as negative energy solutions with energy E = p2 + m2 . How do we interpret these negative energy solutions? The conventional interpretation is called the Feynman-Stuckelberg interpretation : A negative energy solution represents a negative energy particle travelling backwards in time, or equivalently, a positive energy antiparticle going forwards in time. 11

and collecting them together 1 0 0 1 u1 = N1 pz u2 = N2 px ipy E+m

pz and u4 = N4 Em 0 1
pz Em px +ipy Em

Em

(76)

px ipy
Em

With this interpretation we tend to rewrite the negative energy solutions to represent positive antiparticles. Starting from the E < 0 solutions pz px ipy px +ipy u3 = N3 Em 1 0
Em

pz and u4 = N4 Em 0 1

Em

(78)

Here we implicitly assume that E is negative. We dene antiparticle states by just ipping the sign of E and p following the Feynman-Stuckelberg convention. v1 (E, p)ei(Etxp) = u4 (E, p)ei(Etxp) v2 (E, p)ei(Etxp) = u3 (E, p)ei(Etxp) (79) (80)

where E = |p|2 + m2 . Note that v1 is associated with u4 and v2 is associated with u3 . We do this because the spin matrix, Santiparticle , for antiparticles is equal to Sparticle i.e. the spin ips for antiparticles as well and the spin eigenvalue for v1 = u4 is spin-up and v2 = u3 is spin-down. In general u1 , u2 , v1 and v2 are not eigenstates of the spin operator (Check for yourself). In fact we should expect this since solutions to the Dirac Equation are, by denition, eigenstates of the Hamiltonian operator, H, and the Sz does not commute with the Hamiltonian : [H, Sz ] = 0, and hence we cant nd solutions which are simultaneously solutions to Sz and H. However if the z-axis is aligned with particle direction : px = py = 0, pz = |p| then we have the following Dirac states |p| 1 0 0 E+m 0 0 |p| 1 E+m u1 = N |p| u2 = N 0 v1 = N 0 v2 = N 1 E+m |p| 1 0 0 E+m (81) These are eigenstates of Sz 1 Sz u 1 = + u 1 2 1 Sz u 2 = u 2 2 1 anti Sz v1 = Sz v1 = + v1 2 1 anti Sz v2 = Sz v2 = v2 2 (82) (83)

So we have found solutions of the Dirac Equation which are also spin eigenstates....but only if the particle is travelling along the z-axis. 12

0.4.3

Charge Conjugation

Classical electrodynamics is invariant under a change in the sign of the electric charge. In particle physics, this concept is represented by the charge conjugation operator that ips the signs of all the charges. It changes a particle into an anti-particle, and vice versa: C|p >= |p > (84)

. Application of C twice brings us back to the the orignal state : C 2 = 1 and are 1. In general most particles are not eigenstates of so eigenstates of C C. If |p > were an eigenstate of C then C|p >= |p >= |p > (85)

implying that only those particles which are their own antiparticles are eigen states of C. This leaves us with photons and the neutral mesons like 0 , 0 and . C changes a particle spinor into an anti-particle spinor. The relevant operation is C = C 0 (86) with C = i 2 0 . We can apply this to, say, the u1 solution to the Dirac Equation. if = u1 ei(Etxp) , the C = i 2 = i 2 u ei(Etxp) 1 px ipy 1 0 0 0 i E+m 0 0 0 i 0 pz 2 pz E+m i u1 = i 0 i 0 0 N1 E+m = N1 0 = v1 px +ipy i 0 0 0 1 E+m

(87)

or, in full, C(u1 ei(Etxp) ) = v1 ei(Etxp) . Hence charge conjugation changes the particle eigenspinors into their respective anti-particle spinors.

0.4.4

Helicity

The fact that we can nd spin eigenvalues for states in which the particles are travelling along the spin-direction indicates that the quantity we need 13

is not spin but helicity. The helicity is dened as the projection of the spin along the direction of motion: h = p = 2S p = 0 p 0 (88)

and has eigenvalues equal to +1 (called right-handed where the spin vector is aligned in the same direction as the momentum vector) or -1 (called left-handed where the spin vector is aligned in the opposite direction as the momentum vector), corresponding to the diagrams in Figure1.

Figure 1: The two helicity states. On the left the spin vector is aligned in the direction of motions of the particle. This is a right-handed helicity state and has helicity +1. On the right the spin vector is antiparallel to the particle momentum. This is a left-handed helicity state with helicity -1. It can be shown that the helicity does commute with the Hamiltonian and so one can nd eigenstates that are simultaneously states of helicity and the Hamiltonian. The problem, and it is a big problem, is that helicity is not Lorentz invariant in the case of a massive particle. If the particle is massive it is possible to nd an inertial reference frame in which the particle is going in the opposite direction. This does not change the direction of the spin vector, so the helicity can change sign. The helicity is Lorentz invariant only in the case of massless particles.

0.4.5

Chirality

Wed rather have operators which are Lorentz invariant, than commute with the Hamiltonian. In general wave functions in the Standard Model are eigenstates of a Lorentz invariant quantity called the chirality. The chirality operator is 5 and it does not commute with the Hamiltonian. Due to this, it 14

is somewhat dicult to visualise. The best picture I can get comes from the denition : something is chiral if its mirror image does not superimpose on itself. Think of your left hand. Its mirror image (from the point of view of someone in that mirror looking back at you) is actually a right hand. It and its mirror image cannot be superimposed andit is therefore an intrinsically chiral object. In the limit that E >> m, or that the particle is massless, the chirality is identical to the helicity. For a massive particle this is no longer true. In general the eigenstates of the chirality operator are 5 uR = +uR 5 uL = uL 5 vR = vR 5 vL = +vL (89)

, where we dene uR and uL are right- and left-handed chiral states. We can dene the projection operators 1 1 PL = (1 5 ) PR = (1 + 5 ) 2 2 (90)

such that PL projects outs the left-handed chiral particle states and righthanded chiral anti-particle states. PR projects out the right-handed chiral particle states and left-handed chiral anti-particle states. The projection operators have the following properties :
2 PL,R = PL,R ;

PR PL = PL PR = 0;

PR + PL = 1

(91)

Any spinor can be written in terms of its left- and right-handed chiral states: = (PR + PL ) = PR + PL = R + L (92) Chirality holds an important place in the standard model. Lets take a standard model probability current We can decompose this into its chiral states = (L + R ) (L + R ) = L L + L R + R L + R R
1 1 Now, L = L 0 = 2 (1 5 ) 0 = 0 1 (1 + 5 ) = 2 (1 + 5 ) = PR 2 where I have used the fact that 5 = 5 and 5 0 = 0 5 . Similarly

(93)

15

R = PL . Using this, it is easy to show that the terms L R and R L are equal to 0: L R = PR PR 1 1 = (1 + 5 ) (1 + 5 ) 2 2 1 = (1 + 5 )( + 5 ) 4 1 = (1 + 5 )( 5 ) 4 1 = (1 + 5 )(1 5 ) 4 1 = (1 + 5 5 ( 5 )2 ) 4 = 0 (94) (95) (96) (97) (98) (99) (100)

since ( 5 )2 = 1 and 5 = 5 . Similarly for the other cross term, R L . Hence, = L L + R R (101)

So left-handed chiral particles couple only to left-handed chiral elds, and right-handed chiral elds couple to right-handed chiral elds. One must be very careful with how one interprets this statement. What it does not say is that there are left-handed chiral electrons and right-handed chiral electrons which are distinct particles. Useful though it is when describing particle interactions, chirality is not conserved in the propagation of a free particle. In fact the chiral states L and R do not even satisfy the Dirac equation. Since chirality is not a good quantum number it can evolve with time. A massive particle starting o as a completely left handed chiral state will soon evolve a right-handed chiral component. By contrast, helicity is a conserved quantity during free particle propagation. Only in the case of massless particles, for which helicity and chirality are identical and are conserved in free-particle propagation, can left- and right-handed particles be considered distinct. For neutrinos this mostly holds.

16

0.5

What you should know

Understand the covariant form of the Dirac equation, and know how to manipulate the matrices. Know where the denition of a current comes from. Understand the dierence between helicity and chirality. Be able to manipulate gamma matrices.

0.6

Futhur reading

Griths Chapter 7, Sections 7.1, 7.2 and 7.3 Griths Chapter 10, Section 10.7

17